Company Overview
Most investors who’ve heard of Marvell think of it as a legacy semiconductor company making storage and networking chips. That was the old story. The new story — the one that has sent MRVL up more than 120% year-to-date — is about a company that has become one of the two dominant architects of custom AI silicon for the world’s largest hyperscalers. Marvell doesn’t sell chips the way Nvidia does; it designs made-to-order AI accelerators and high-speed interconnect products for companies like Amazon, Microsoft, and Google that want chips built specifically for their own AI workloads. And on March 31st, Nvidia validated that business in the most concrete way imaginable — by writing Marvell a $2 billion check.
The Nvidia investment was structured as a strategic partnership around NVLink Fusion, Nvidia’s platform that allows custom third-party accelerators to plug directly into its own AI infrastructure ecosystem. Marvell will supply custom XPUs and optical interconnect technology; Nvidia will provide everything surrounding them. The deal essentially co-opted Marvell into the Nvidia ecosystem rather than leaving it to compete against it. Tomorrow morning, Marvell reports fiscal Q1 FY2027 earnings — and a wave of fresh analyst price target increases this week, with targets ranging from $195 to $220, signals that Wall Street expects the momentum to continue.
Key Technical and Fundamental Drivers
Nvidia’s $2 Billion Strategic Investment → The Clearest Possible Endorsement On March 31st, Nvidia invested $2 billion in Marvell via convertible preferred stock and entered a formal partnership around NVLink Fusion. The deal places Marvell’s custom XPUs, optical DSP, and silicon photonics directly inside Nvidia’s AI rack-scale architecture — a structural relationship that strengthens both companies’ positions in hyperscaler data centers simultaneously. RBC analyst Srini Pajjuri said it plainly: the investment is “meaningful validation of Marvell’s dominance in optical connectivity infrastructure.”
AWS Trainium → Largest Custom Silicon Program in the Industry Marvell’s single biggest revenue driver is its design work for Amazon’s Trainium AI accelerators — the chips AWS uses to train large language models internally and offer to customers as a GPU alternative. Marvell’s relationship with Amazon extends through Trainium3 in 2026 and into Trainium4, which is being built with NVLink Fusion compatibility. Wells Fargo’s May 20th upgrade to $195 anchored specifically on this Trainium deployment ramp as the primary near-term driver.
Pre-Earnings Wave of Upgrades → Five Firms Raised Targets This Week In the days since Nvidia’s earnings, five major Wall Street firms raised Marvell price targets simultaneously: Melius Research to $220, Citigroup to $215, Bank of America to $200, RBC Capital to $200, and Wells Fargo to $195. Bank of America’s Vivek Arya named MRVL a top semiconductor pick alongside Nvidia, Broadcom, Micron, and AMD — a group that gets far more retail attention than Marvell typically does. The cluster of upgrades ahead of tomorrow’s print signals broad institutional conviction.
Record Revenue → $2.075 Billion, Up 37% Year-Over-Year Marvell’s most recent quarter showed record revenue of $2.075 billion, with data center revenue of $1.518 billion representing 73% of the total. CEO Matt Murphy guided for full-year FY2026 revenue growth exceeding 40%, with data center growth expectations raised above prior targets. Analysts are projecting Q1 FY2027 EPS of $0.79, implying 27% year-over-year growth, on revenue of approximately $2.40 billion.
Celestial AI Acquisition → Silicon Photonics for the Next Wave In early 2026, Marvell acquired Celestial AI, a silicon photonics company that adds optical fabric technology to its portfolio. The acquisition was one of the reasons Nvidia chose Marvell as its NVLink Fusion partner — the combined capability in custom XPUs plus optical interconnect is positioned to power the semi-custom AI rack reference designs Nvidia and Marvell plan to bring to hyperscalers targeting next-generation Rubin GPU clusters by Q3 2026.
Market Takeaway
Tomorrow’s earnings report is the next concrete test of whether Marvell’s remarkable 2026 rally is built on durable fundamentals or has run ahead of itself. The bull case is straightforward: custom AI silicon is the fastest-growing segment in semiconductors, Counterpoint Research estimates global AI ASIC shipments will triple between 2024 and 2027, and Marvell sits alongside Broadcom as one of only two companies capable of designing the most complex custom accelerators at hyperscaler scale. The Nvidia $2 billion investment removed the narrative risk that Marvell was competing against the dominant platform — instead it’s now integrated into it.
The risk worth watching is valuation. At over 40 times forward earnings and with the stock up 120% year-to-date, there’s little room for a guidance miss or any softening in data center commentary. Several analysts note that near-term upside on Trainium could be constrained by tight wafer supply at TSMC, and some insider selling has accompanied the rally. For traders watching the setup, tomorrow’s print offers a live test: a guidance raise and accelerating custom silicon revenue numbers would confirm the thesis the five Wall Street upgrades this week are banking on. A print that merely meets expectations on a stock priced for perfection could be a different kind of story — one worth watching the tape carefully on both sides.